Fuelly

[Darlo]

MPG =
378
- 0.000033 Aggressive cubed
+ 0.00430 Aggressive squared
- 0.158 Aggressive %
- 0.000401 A/C squared
+ 1.39 Cutting engine at lights
- 0.000882 city X aggressive
- 0.647 Fills since oil change
+ 0.0372 fills since oil change squared
- 0.000063 pounds of gas cubed
+ 0.0335 pounds of gas squared
- 5.85 pounds of gas on average

I have the variables:
% Aggressive driving in a given tank(then squared and cubed of variable)
% A/C use in a given tank (then squared too)
Tanks of gas that I turned off the car at lights (either I did or did not, not a percentage)
# of fills since last oil change (average fill is 400 miles)
pounds of gas in my tank on average in a given tank (squared and cubed)

I also have a cross product in there as well (city X aggressive) which takes my %city driving and multiplies the % of time in that same tank that I drove aggressively.

I had 23 data points (23 fills over the course of a few months).
This regression model gives a R-squared value of 92.9%. (which means that 92.9% of the variability in my gas mileage is accounted for in this equation.)

Oh and the reason I added the pounds of gas in the tank is due to the debate over whether getting gas too often (aka when you still have half a tank) will hinder your mileage. And it does indeed seem to.

(All percentage values were timed by 100)

As for aggressive driving, I defined it as the percent of the miles driven that I
a) would speed (to a maximum of 10 mph above the limit)
b) purposely accelerate and brake needlessly (aka in the city in between lights)
c) drive in a lower gear (aka higher revs for same speed)
And this percentage may be off by 5% or so either way, but as per the large "n" normality in the data's variation would indeed even out.

[Pu241]

OK, looked this over and I'm not a "stats guy" so can you give me a rough accounting of the what numbers below 378 actually mean.

[Pu241]

MTec007

what did you do to get the 49.7 MPG on the 10/16 fill up?

[Darlo]

OK, so every time I filled up the car, I recorded my miles travelled, gallons used, what % of the time I used the A/C, what % of the time I drove aggressively (as per my own definition), whether or not I cut my engine at lights and drive-thru's, kms since my last oil change, and the % city driving I did during that tank of gas.....

All this went into a spreadsheet, and then I made some extra columns that had the value of aggressive driving squared, then cubed in another column, the in yet another column A/C % squared, cubed etc etc....... The purpose of this was to observe not only the linear decrease of MPG with the linear increase of aggressive driving, but to also account for any squared relationship, or cubed relationship.

I then used a stats program called Minitab to do a "least squares analysis" (which, after a bunch of work on my part, finds what variables were important in predicting gas mileage, and which were not) The variables that were not a good predictor of gas mileage were removed and the model was rerun (this is done one variable at a time, and there are many conventions about what order this should be done....I took an upper level stats course, so dont feel too bad if this makes no sense, lol)

In the end, if I were to fill up that same car now, and plug into the model above the aggressive driving I did, and the A/C I used etc, it will spit out a MPG estimate which will be VERY accurate to the calculated value that you would normally use when you go to the pump...........

One thing that this model also tells you, if I posted the p-value for each variable, you would see what variables were most important, or weighted heavier in the prediction of MPG. (lower p-values are best)

Dont worry if this makes no sense, just remember these results.....

1) Aggressive driving is the biggest factor (by a lot)
2) Cutting your engine at lights gave me 1.39 more mpg on average
3) Getting regular oil changes did indeed help (only a fraction of a mpg)
4) A/C usage did decrease mpg, but by only about a 1/4 mpg.
5) City driving was the worst when coupled with aggressive driving habits
6) Filling the tank when it is only half empty will decrease your mpg because of the extra weight of the other half tank that you are basically just carting around for no reason......

[Darlo]

For the stats people on the site, I welcome any help in explaining multiple regression to the non-math people, lol......

[Darlo]

Does that make more sense Pu241?

[Pu241]

Yes, helps alot.
Was going to take the MiniTab class, but I had a coworker who literally could look at a list of numbers and tell you its standard deviation.
I thought why do I need to learn it when I have him around!
Not to mention he was the MiniTab guru too.

I presume the same technique can be used to determine the most effective aero mods?

[ericsmith32]

I'm going thru and entering a majority of the fill ups since 2005. Dad says he gets a little better mileage than I and on occasion he has. But the majority the average is the same. I might enter in a couple years worth from his other one just to compare. You'd think that the 91 basic L would get a little better than the 91 GL but I don't think so.

[Darlo]

Yes the same thing can be used to determine the effectiveness of the mods......At the time I was more interested in things "normal" people would do to increase mpg.....

[ericsmith32]

Hmm it does come in handy to check and track mileage. It looks like the mileage started to drop around last Oct. Surprised he didn't mention it. Pretty sure it was the rear wheel bearings.. the grease looked like curdled cottage cheese. Although the bearings looked new. Cleaning the cap and rotor and replacing the Bosch o2 just in case though.

[Pu241]

Another tank emptied and fuelly gains another data point!
3 work days per tank!
This post doesn't yet show the fuelly update to 6 tanks burned ( 7:35PM).

OK, just takes some time to work it way through (7:48PM).

[Darlo]

Well done! My mpg has gone down as the weather hear has been really rainy, and I drove up a mountain, lol..... Only got 46mpg embarrassing :p

[Darlo]

Just removed my side mirror, as to reduce the drag! Hope it works!

[Pu241]

It will reduce drag, but you might not see an increase in MPG as it might be "below the level of noise".
Also believe you previously stated you seldom exceed 50 mph, so air resistance is somewhat reduced in that realm.
You going to use a wider inside mirror?
Or some other means of seeing behind you on that side.
Wonder if you can remove both pass and drivers side mirrors for double effect.
But of course, that is illegal.
If they catch you!

[ericsmith32]

Done! I guess almost 41 isn't to shabby. Dad was right about winter mileage. This one drops almost 6 mpg. Not sure if it the TB coolant bypass that was causing that but I switched it back yesterday.. probably didn't help.